3236 IEEE TRANSACTIONS ON MAGNETICS,VOL. 30, NO.
5, SEPTEMBER 1994
Modelling of Electromechanical Relays taking into account
Movement and Electric Circuits
F. Henrotte, A. Nicolet, H. Hkdia, A. Genon, W. Legros
University of Liege  Dept of Electrical Engineering
Institut Montefiore  Sart Tilman Bit. B28  B4000 Liege (Belgium)
Abstract  This paper presents a numerical subdomain Di. Both methods are based on spatial
modelling of an electromechanical relay connected discretization relations such as :
with an electric excitation circuit. This transient
N
modelling not only takes into account the classical
electromagnetic equations of the device but also the A(rr,t) = Ak(t) wkb)
k= 1
movement and circuit equations. The use of the M (2)
finite elementboundary element coupling method
facilitates the computation of the movement while Htkd = Htk(t) Wkb)
k= 1
the actual coupling with circuit equations is
necessary for an accurate and reliable where wkQ are nodal weighting functions, Ak are the N
representation of transient phenomena. time varying nodal values of the vector potential and Htk are
the M time varying nodal values of the magnetic tangential
INTRODUCTION
field on the boundaries Ti.
In the majority of practical applications, electromagnetic The finite element method with trial functions,
phenomena are used to cause movement or forces (motors, { Wn(2L),i=l,...,NI, identical to the weighting functions
relays, ...) and are inserted in external electric circuits (Galerkin method) gives N equations such as (3).
( transformers, ... ). Magnetic field programs that are only
able to treat electromagnetic quantities are therefore not very
effective in numerical modelling of such apparatus. Moreover
the coupling is often strong in well designed systems which
are precisely conceived to transform one kind of energy into
I i
( Vi grad A.grad Wn  Ji Wn ) dDi 
In a subdomain made of a linear and nonconducting
material, it is also possible to use the boundary element
I,
Wn H t dTi = O(3)
mother (e.g. a motor has to transform electrical energy into method which is based on equation (4) where G is the free
nechanical energy ). In a relay for example, the excitation space Green function for 2D Laplace operator, c is a constant
circuit generates current in the coil and the resulting magnetic depending on the smoothness of Ti (c=1/2 on a regular
field causes force and torque on moving parts. In the reversed boundary) and n is the outer normal vector for Di.
way, motion of moving parts acts on geometry, induces flux
variations in the system and modifies the inductance in the
circuit.
I MAGNETIC FIELD MODELLING
To compute the electromagnetic part of the problem (and
to evaluate forces), a program called LUCIE is used which is
based on the coupling of the finite element method and the magnetic circ:uit
boundary element method (F.E.M.B.E.M. coupling).
(FEW
a. Subdomain method
To carry out that coupling, the system is divided into N
subdomains (Di, i=l, ... ,N) whose respective boundaries
are the contours (Ti, i=l, .,.,N}. The unknown fields are
the vector potential A(L,t) on the subdomains Di and the
tangent magnetic field H&t) on the boundaries Ti. Under
these conditions, the 2D magnetic field equation (1) (which is
a set of partial differential equations) is well posed in each
subdomain Di.
P

air (BEM)
30"
div (Vi grad A) = .Ti (1)
Fig. 1. Representation of the subdomains defined in the modelling
Either the finite element method (F.E.M.) or the boundary of the relay. The solving method is specified for each subdomain.
element method (B.E.M.) can be used to solve (1) in each
b. F.E.M.B.E.M. coupling
Manuscript received November 1, 1993. is to use
This text presents research results of the Belgian programme on
The philosophy Of the F.E'M*B+E*M*
Interuniversitv Poles of Attraction initiated bv the Belgian State. Prime the B.E.M. for linear magnetic and non conducting media
Minister's Office, Science Policy Programming.The Scientific wherein rigid parts can move (generally air), remeshing is
responsibility is assumed by its authors.
00189464/94$4.00 Q 1994 IEEE
3237
therefore avoided. The B.E.M. is also used to provide a where nx and ny are the components of the outer normal
rigorous treatment for open boundaries. The F.E.M. is vector n, 4 is the local coordinate of the shape functions used
chosen everywhere else and especially in media that can be for the boundary element discretization and J(4) is the
the seat of phenomena such as saturation or eddy currents. corresponding Jacobian.
11 COUPLING WITH MOVEMENT [ l ] c. Coupling with the magnetic system
Due to the fact that the expression of the force is an
a Mechanical model intricate nonlinear function of the magnetic degrees of
Only rigid motions in 2D models are considered. Three freedom, it is cumbersome to solve simultaneously the finite
degrees of freedom are associated with each moving part, two element equations and the mechanical equations.
degrees of translation and one degree of rotation. The Consequently, the following algorithm is adopted to model
mechanical model is the classical Newton equation taking the movement :
into account electromagnetic forces and torque. Connection of
the different objects with springs, viscous dampers or Compute electromagnetic state at time t
constant forces is considered. Compute the magnetic force and torque for this state
The movement fmt order differential equations are : Solve the mechanical equations to find the
dv displacements of the moving parts.
m+cv+kx=F(t) , v= dx Verify coherence:
dt dt
where the following quantities are defined : Keep position in physical limits
position : x(t); Modify time step if necessary
speed :v(t); Modify the geometry in the database using the computed
displacements
mass or momentum of inertia : m;
Next step
elastic constant of the spring : k;
viscous damper constant :c;
This kind of algorithm can be qualified as "weak
applied force or torque including magnetic forces and coupling" because the magnetic and mechanical equations are
constant forces: F(t). not solved simultaneously; that means that the magnetic
force is supposed to be a constant on one time step.
Time discretization by the backward Euler method gives :
v(t+At) = mv(t) + F(t)At  k x(t)At
111 COUPLING WITH CIRCUITS [ 2 ]
ut2+ cAt+ m (6)
From an electrical point of view, the conductors of a
mx(t) + mv(t) At + F(t)Atz + cx(t)At magnetic field model can be seen as self and mutual
x(t+At) = inductances with resistances. An approach for coupling them
Ut2+ cAt + m with an external excitation circuit could be to extract an
b. Computation of magnetic forces equivalent impedance matrix from the numerical model and to
introduce it in the electrical circuit equations. Unfortunately,
The F.E.M.B.E.M. coupling method gives, at each time because of saturation, movement and skin effect, that matrix
step, not only the knowledge of the vector potential field A would depend on the level of excitation, on the position and
but also the knowledge of the tangential magnetic field Ht on on the frequency. That makes it necessary to realize a real
the boundaries r i . That allows an easy computation of coupling of the finite element equations with the circuit
magnetic forces by the Maxwell stress tensor method [3]. equations.
That method gives an equivalent normal force density Fn and
an equivalent tangential force density Ft on the boundaries a. Edcj, currents
of objects whose expressions are : In transient problems where eddy currents exist in the
conductors, the current density term J(rr,t) in equation (2) is
F = ( W*  yO H t ) n , Ft =B,H,t (7) not known a priori and is then replaced by its expression in
accordance with Ohm's law :
Note that the quantities Bn and Ht which are involved in aA
those expressions are well defined even on surfaces of at
J i = o i E =  ai( + Ui 1. (9)
discontinuity ; the scalar field Ht is directly given by the The conductor, Di, is characterized by its conductivity Oi.
coupling method while Bn is computed from the vector The time partial derivative m a t stands for the inductive
potential by : effect and Ui is the terminal voltage p.u. of length imposed
on the conductor.
A crucial problem is to determine the effect of movement
on eddy currents. According to Faraday's induction law, the
induced electromotive force in a circuit is equal to the total
magnetic flux variation through the circuit whatever the
3238
origin of the variation : time variation of the field or
movement of the circuit. Equation (9) is therefore only IV EXAMPLE : MODELLING OF THE RELAY
applicable for non moving conductors and the correct
expression for a conductor moving at speed V i is [41: a. Movement equations
aA In this problem, only one direction of movement is free
Ji = ai(E+ vi x B) = ~i(+Ui  vi x B) (10) and there is only one movement equation. The moving core
at (mass=0.2 kg) is fKed to a spring (k=2500. N/m) tending to
where B is the induction. Since Vi x B = (vi.grad) A (as keep the relay open while the magnetic force tends to make
it may be verified component by component), the the closure (Fig. 2). The movement is 7mm long; the relay
generalization of (9) for moving conductors is : is 5 0 " deep; friction is neglected.
dA
Ji =  Oi( dt + Ui ) (1 1)
where d/dt is the material derivative, i.e. the variation for a
given particle. We have pointed out before that with the Fig. 2. Mechanical diagram of the relay.
F.E.M.B.E.M. coupling method the mesh of moving parts
is involved as a whole in the movement. Each node of the b. Circuit equations
system corresponds to the same particle during all the The excitation circuit (Fig. 3) consists of an initially
motion. The material derivative dNdt is simply evaluated by charged capacitor C (Vc = 1OV) connected with an extemal
difference of nodal values; there is no need for an explicit resistance R and with the two conductors in series
term for the motional induced electromotive force. representing the coil in the model (Nc = 2). The capacitor
b. Erteml degrees offeedomfor conductors voltage has to be chosen as a state variable (Ne = 1). The
The magneticelectrical coupled problem can be solved if circuit theory gives the following three equations :
an imposed terminal voltage is given for each conductor of VcRII +U1U2=0
the problem. If the total current, Ii, is given instead of the I1 + I 2 = 0 (14)
terminal voltage, this becomes one more unknown and one 11+C dVC = O
more relation is necessary to relate it to the magnetic degrees dt
of freedom. Since the tangential field Ht is involved as an
unknown in the F.E.M.B.E.M. coupling, the following
expression of Ampere's law gives such a relation :
C R
Ii= Ht dTi (12)
ri
A special treatment is possible to model coils where eddy
currents are prevented by the smallness of the wire section. In 1' I
that case, the following relation is used instead of (12) [l]:
NiIi =  Oi
d dt
dDi  OSiUi
where Ni is the number of tums and Si the section of one
(13)
turn in the subdomain Di. The current density is given by
Fig. 3. Electrical diagram of the relay. The conductors are
Ji = Ii/Si instead of (11) and there is no more skin effect, represented by their external degrees of freedom Vi and 1;.
only a global inductive effect is allowed.
From a more general point of view, we can systematically V RESULTS
relate to every conductor of the problem an equation such as
(12) or (13) and two extemal degrees of freedom: the total As a deliberate choice for this modelling, the electrical and
current in the conductor, Ii, and the terminal voltage, Ui . mechanical diagrams are extremely simple when considered
One circuit equation is necessary for each conductor to solve separately (Fig. 2 and 3). They present elementary and well
the system. The simplest case corresponds to imposing Ui or known individual behaviours.
Ii for each individual conductor but we are more interested in In spite of this, time evolutions become quite intricate as
the general case of Nc independent conductors connected with soon as the coupled problem is considered and a global
any extemal R.L.C. circuit according to any topology. analysis of the results is not of interest. We rather apply
If that extemal circuit has Ne state variables (the number ourselves in this paper to demonstrate successively that
of inductances and capacitors), the circuit theory gives Nc+Ne expected interplays appear actually in computed results.
relations or first order differential equations which are added to Two figures are presented showing time evolutions of the
the magnetic model ones and solved in the same time. total applied force on the moving core (including the
magnetic force and the spring force), the position of the core
3239
and the current in the coil. All curves are scaled down for seen as an RLC device with a kind of damped oscillating
comparison, scaling values are given in Table 1. behaviour.
Fig. 4 and 5 correspond respectively to the cases where Because the inductance depends on the motion, the circuit
the magnetic circuit is made of a linear or a nonlinear is a true RLC circuit as long as the moving core remains
material. Both materials are identical at low fields. motionless. One can check in Fig. 4 that the current curve is
0
a piece of a damped sinusoidal function in that case. On the
other hand this is no longer true in Fig. 5 because of the
nonlinearity of the material.
0 c. Influence of movement on current
0
The downward motion of the moving core decreases the
length of the air gap and increases its reluctivity in the
0 magnetic circuit. An inverse current is then induced in the
coil (in accordance with the Lenz law) which opposes the
0
setting of the direct current. That inverse current reaches a
maximum when the speed does and it vanishes when the
closure is done.
Fig. 4. Time evolutions of the total force, the position of the
moving core and the current in the linear case CONCLUSION
It is obvious that the correct waveform for current in a
relay results from a complex interplay between the electric
circuit and the variation of inductance due to displacement of
moving parts.
A numerical implementation has been done with the
F.E.M.B .E.M. coupling method. This method presents
several advantages for that kind of modelling :
There is no mesh deformation during the movement.
There is no need for an explicit term for the motional
induced electromotive force.
The method makes it easy to evaluate force and torque
by the Maxwell stress tensor method.
The method makes easy it to define the electrical
Fig. 5. Time evolutions in the nonlinear case.
external degrees of freedom of the conductors of the
model and to make the connection with an electrical
linear nonlinear circuit.
I FmaxO 27.32 10.92 Analysis of the computed results shows that they are in
1 Imax(A) I 6.186 6.293 I good agreement with the different physical phenomena
involved in the problem.
Table 1. Scaling values of the total force and the current.
REFERENCES
a. Influence of current onforce
111 A. Nicolet, F. Delinc.4, A. Genon, W. Legros, Finite element
The magnetic force is closely related to the square value of boundary element coupling for the movement modelling in two
the flux crossing the air gap which is itself proportional to dimensional structures, J. Phys. 111 France 2 (1992) pp 20352044.
the current in the coil in the linear case. On Fig. 4 and 5, it P I A. Nicolet, F. Delin&, N. Bamps, A. Genon, W. Legros, A coupling
between electric circuits and 2D magnetic field modelling, IEEE
can be checked that the current and force curves exhibit Transactions on Magnetics, Vol. 29, n2, pp 16971700, March 1993.
similar shapes. They both present the characteristic [31 B.Aldefeld, Forces in Electromagnetical devices,
retrogression point when the closure of the relay occurs, i.e. Compunmag(l978). Grenoble, pap 8.1.
141 L. Landau and L. Lifschitz, Electrodynamique des milieux
when the motion induced current vanishes. continus, Editions Mir, Moscow, 1969.
b. Influence of the electric circuit on the magnetic model 151 B. Lequesne, Dynamic model of solenoids under impact excitation,
including motion and eddy current, IEEE Trans. Mag., Vo1.26 No.2,
As pointed out before, the magnetic model can be seen as pp.11071116, March 1990.
161 Y.Kawase and Y.Ohdachi,Dynamic analysis of automotive solenoid
an RL device if Ui or Ii is imposed on each conductor. valve using finite element method, IEEE Trans.Mag., Vo1.27, pp.
Resulting time evolutions are combinations of exponential 393942, 1991.
functions. [71 E.Vassent,G.Meunier,A.Foggia and G.Reyne, Simulation of
inductive machine operation using a step by step finite element
The situation is completely different when the conductors method coupled with circuits and mechanical equations, IEEE
are inserted in an electric circuit, especially if this one Trans. Mag. Vo1.27, pp.523234, 1991.
contains capacitors. In this case, the system must rather be 181 S.J. Salon, Coupling of transient fields, circuit and motion using finite
element analysis, Proceedings of IMACS91, Dublin, July 1991.